Ross Taylor Profile picture
Building @GenReasoning. Previously lots of other things like: Llama 3/2, Galactica, Papers with Code.
Sep 17, 2025 14 tweets 4 min read
Supplementary information for the new DeepSeek R1 Nature paper is very interesting!

Details on training data, hyperparameters, base model importance, and more. Image static-content.springer.com/esm/art%3A10.1…